Large scales are more predictable

For a realistic NWP model the range of meteorological scales is the same throughout the forecast; a D+10 forecasts looks like an analysis of the atmosphere: all scales are represented in a realistic way. But most of these realistic looking features in the D+10 forecast will of course not verify. They will be in the wrong place, with the wrong intensity and mostly not exist at all.

1. The relation between the scale of an atmospheric feature at 500 hPa, measured by the number of spectral components needed to describe it, and its predictability. Forecasts beyond six days normally contain useful information only in the first 10-15 spectral components. Since 1993 the predictability measured in this way has increased by two days.

The predictability of atmospheric motion scales decreases rapidly throughout the forecast, starting with the smallest scales. Small baroclinic systems or fronts are well forecasted up to around D+3, cyclonic systems around D+5 and the long planetary waves up to around D+10.

By relying on his experience of what is normally predictable at a certain range, the forecaster using the ECMWF deterministic forecasts can disregard the small and unpredictable scales, and concentrate on the large and predictable scales Doing so he will most of the time be able to make useful forecasts on average up to a week ahead.